Evolutionary Multi-Task Optimization: Foundations and Methodologies

The human brain's ability to manage multiple tasks simultaneously and use knowledge from one task to improve problem-solving in other related tasks is called cognitive multitasking. Machine learning has explored leveraging relevant information across related tasks as inductive biases to enhance learning performance, while attempts to emulate the human brain's generalization in optimization have received less attention. Evolutionary Multi-Task (EMT) optimization is a novel evolutionary search paradigm that conducts searches concurrently on multiple search spaces corresponding to different tasks or optimization problems, exploiting latent synergies among distinct problems. This book discusses the foundations and methodologies of developing evolutionary multi-tasking algorithms for complex optimization, including domains with multiple objectives, high-dimensional search spaces, and NP-hardness.

Format: Hardback
Length: 219 pages
Publication date: 30 March 2023
Publisher: Springer Verlag, Singapore

The human brain's remarkable ability to handle numerous tasks simultaneously is a testament to its extraordinary complexity. The knowledge acquired from one task can be applied to improve problem-solving in other related tasks, showcasing the interconnectedness of our cognitive abilities. In the realm of machine learning, the concept of leveraging relevant information across related tasks as inductive biases has gained significant attention. This approach aims to enhance learning performance by capitalizing on the shared characteristics and patterns among tasks.

On the other hand, attempts to replicate the human brain's ability to generalize in optimization, particularly in population-based evolutionary algorithms, have received relatively limited attention. However, recent developments in evolutionary computation have introduced a novel evolutionary search paradigm known as Evolutionary Multi-Task (EMT) optimization. Unlike traditional evolutionary searches that focus on solving a single task in a single run, EMT algorithms conduct concurrent searches across multiple search spaces, each corresponding to a different task or optimization problem. These search spaces possess distinct function landscapes, allowing for the exploration of synergies among different problems.

By harnessing these latent synergies, EMT optimization has demonstrated exceptional search performance in a wide range of continuous, discrete, and hybrid (mixture of continuous and discrete) tasks. This breakthrough has significant implications for solving complex optimization problems, particularly those characterized by multiple objectives of interest, high-dimensional search spaces, and NP-hardness.

This book aims to provide a comprehensive exploration of the foundations and methodologies for developing evolutionary multi-tasking algorithms. It delves into the theoretical underpinnings of EMT optimization, discusses its advantages over traditional approaches, and presents practical examples and case studies from various domains. By presenting a comprehensive understanding of EMT optimization, this book serves as a valuable resource for researchers, practitioners, and students interested in advancing the field of machine learning and optimization.

Weight: 518g
Dimension: 235 x 155 (mm)
ISBN-13: 9789811956492
Edition number: 1st ed. 2023